#Join the data boards
I loaded september to october data boards from 2017 - 2023 to represent the freshers season then I merged the health boards #I summarise total antidepressant prescriptions per Freshers year and plotted the graph to see the trend #Looked at pre coviid , during covid and after covid trend to see if there has been any impact or association
library(tidyverse)
library(here) # directory stucture
library(gt) # tables
library(janitor) # cleaning data
library(ggplot2) # plotting graph
library(sf) # to read in map data
library(readxl) # to read in map data
library(plotly) # to make interactive
library(viridis)
library(sf)
loading a large amount of data in a shorter time period by downloading and using the mapdfr function (data from 2017-2023)
files <- list.files(here("data", "winter_data"), pattern = "csv")
winter_data <- files %>%
map_dfr(~read_csv(here("data", "winter_data", .))) %>%
clean_names()
clean up data and filter for the sections you want
filtered_winter_data <- winter_data %>%
filter(str_starts(bnf_item_code,"0403")) %>% #antidepressant code is 0403
mutate(year = as.numeric(substr(paid_date_month,1,4)), month = as.numeric(substr(paid_date_month,5,6))) %>% #separates the date into years and month so that i can group winter sections
mutate(winter_year=case_when(month == 12 ~ year + 1,
month %in% c(1,2) ~ year) )#makes a new column to group the winter years
filtered_winter_data <- filtered_winter_data %>%
unite("healthboards",hbt2014,hbt,sep = "_")#so some of my data healthboard codes were under the name hbt_2014 AND another was hbt so i had to merge the column so all the healthboard columns fall under one
filtered_winter_data$healthboards <- gsub("[NA]","",filtered_winter_data$healthboards)
filtered_winter_data$healthboards <-
gsub("_","",filtered_winter_data$healthboards)#had to remove some NA characters and '_' characters
Graph 1
winter_years_data <- filtered_winter_data %>%
group_by(winter_year) %>%
summarise(total_items=sum(number_of_paid_items,na.rm = TRUE))
plot <- ggplot(winter_years_data,aes(x=winter_year,y=total_items)) +
geom_line(linewidth=0.7,colour = "blue") +
geom_point(size=2)+
scale_x_continuous(breaks=2017:2023) +
labs(title="Antidepressant Prescriptions During Winter Season",x="Year",y="Total Antidepressant Prescriptions") +
theme_minimal()
ggplotly(plot)
print(plot)
#write a code talking about the zoomed in changes and reference why you dudnt go from 0
#ROUGH
population <- readxl::read_excel(here("data","population.xlsx"), skip=10) %>%
clean_names() %>%
group_by(x2,all_people) %>%
summarise () %>%
filter(!is.na(all_people))
population <- population %>%
rename(h_bname = x2)
filtered2_winter_data <- filtered_winter_data %>%
group_by(healthboards,bnf_item_code,paid_quantity,winter_year,gp_practice) %>% summarise(total_paid = sum(paid_quantity, na.rm =TRUE))
SIMD <- readxl::read_excel(here("data","SIMD.xlsx")) %>%
clean_names() # loading excel data
filtered_SIMD <- SIMD %>%
group_by(simd2020v2_quintile,h_bcode,h_bname) %>%
summarise()
filtered_SIMD <- filtered_SIMD %>%
rename(healthboards = h_bcode)
Overall_SIMD_winter <- filtered2_winter_data %>%
full_join(filtered_SIMD,by = "healthboards")
'relationship = "many-to-many"'
## [1] "relationship = \"many-to-many\""
Overall_SIMD_population_winter <- Overall_SIMD_winter %>%
full_join(population,by="h_bname")
antidepressant_per_head <- Overall_SIMD_population_winter %>%
group_by(healthboards,h_bname,winter_year) %>%
summarise(quantity_per_head = sum(paid_quantity)/mean(all_people))
### map
NHS_healthboards <- st_read(here( "data", "Week6_NHS_HealthBoards_2019.shp")) %>%
clean_names() %>%
rename(h_bname = hb_name)
## Reading layer `Week6_NHS_healthboards_2019' from data source
## `/Users/olufimihanfaturoti/Year 3 Medicine/data-science/B251495/data/Week6_NHS_healthboards_2019.shp'
## using driver `ESRI Shapefile'
## Simple feature collection with 14 features and 4 fields
## Geometry type: MULTIPOLYGON
## Dimension: XY
## Bounding box: xmin: 7564.996 ymin: 530635.8 xmax: 468754.8 ymax: 1218625
## Projected CRS: OSGB36 / British National Grid
# Join spatial data with falls_admissions_75_summary
mapped_data <- antidepressant_per_head %>%
full_join(NHS_healthboards,by="h_bname") %>%
st_as_sf()
#CLAUDE :
library(ggiraph)
plot_map <- mapped_data %>%
ggplot() +
geom_sf_interactive( # Changed from geom_sf
aes(fill = quantity_per_head,
tooltip = paste0(h_bname,
"\nWinter Year: ", winter_year,
"\nQuantity per Head: ", round(quantity_per_head, 2))),
colour = "white",
size = 0.1
) +
scale_fill_distiller(palette = "Blues", direction = 1,
name = "Items per Head") +
labs(
title = "Antidepressant Prescriptions per Head",
subtitle = "By Health Board and Winter Year"
) +
facet_wrap(~ winter_year) +
theme_void() +
theme(
strip.text = element_text(size = 12, face = "bold"),
plot.title = element_text(face = "bold", size = 16),
plot.subtitle = element_text(size = 10)
)
interactive_map <- girafe(ggobj = plot_map) # Changed from ggplotly
interactive_map
CHAT
#summarise totals per winter year per practice
winter_year_summary <- Overall_SIMD_population_winter %>%
group_by (healthboards, gp_practice, winter_year) %>%
summarise(total_paid = sum(paid_quantity, na.rm = TRUE))
# average across winter years
winter_year_average <- winter_year_summary %>%
group_by(healthboards,gp_practice) %>%
summarise(avg_paid_over_winters = mean(total_paid, na.rm = TRUE))
#SIMD
winter_with_simd <- winter_year_average %>%
full_join(filtered_SIMD, by="healthboards") %>%
filter(!is.na(simd2020v2_quintile))
CHAT2
box <- ggplot(winter_with_simd,
aes(x=factor(simd2020v2_quintile),
y=avg_paid_over_winters,
fill=factor(simd2020v2_quintile))) +
geom_boxplot(outlier.shape = 21,
outlier.size = 1.5,
outlier.stroke = 0.5,
linewidth = 0.8,
colour = "black",
alpha = 0.7) +
scale_y_continuous(labels = scales::label_number())+
scale_fill_viridis(discrete=TRUE, alpha=0.9) +
geom_jitter(color='red',size=0.4, alpha=0.4) +
theme(
panel.background = element_rect(fill = "lightblue",
colour = "lightblue",
size = 0.5)) +
labs(
title="Average Winter Antidepressant Prescriptions per Practice by SIMD Quintile",
x="SIMD Quintile (1 = Most Deprived)",
y="Avg prescriptions per practice (winter seasons)")
ggplotly(box)
PERCENTAGE CHANGE
SIMD_winter_summary <- winter_year_summary %>%
full_join(filtered_SIMD,by = "healthboards")
'relationship = "many-to-many"'
## [1] "relationship = \"many-to-many\""
SIMD_winter_final <- SIMD_winter_summary %>%
mutate(period =ifelse( winter_year < 2020, "pre","post")) %>%
group_by(simd2020v2_quintile,h_bname, period) %>%
summarise(mean_total = mean(total_paid, na.rm = TRUE), .groups = "drop") %>%
tidyr::pivot_wider(
names_from = period,
values_from = mean_total
) %>%
mutate(
pct_change = ((post - pre) / pre) * 100) %>%
mutate(h_bname = reorder(h_bname, simd2020v2_quintile))
SIMD_winter_final <- SIMD_winter_final %>%
filter(!is.na(pct_change))
chat3
library(ggplot2)
library(dplyr)
lollipop <- SIMD_winter_final %>%
ggplot(aes(
y = reorder(h_bname, -simd2020v2_quintile), # or reorder(h_bname, pct_change)
x = pct_change
)) +
# stick from 0 → % change
geom_segment(
aes(x = 0, xend = pct_change, yend = h_bname),
linewidth = 1,
colour = "grey40"
) +
# lollipop dot
geom_point(
aes(x = pct_change),
size = 2,
colour = ifelse(SIMD_winter_final$pct_change > 0, "blue", "red")
) +
# percent labels on dots
geom_text(
aes(label = paste0(round(pct_change, 1), "%")),
hjust = ifelse(SIMD_winter_final$pct_change > 0, -0.2, 1.2),
size = 3
) +
scale_x_continuous(labels = function(x) paste0(x, "%")) +
labs(
title = "Percentage Change in Antidepressant Prescriptions",
subtitle = "Lollipop plot (COVID Post vs Pre), ordered by SIMD",
x = "Percentage Change",
y = "Health Board"
) +
theme_minimal(base_size = 13)
ggplotly(lollipop)
questions : not sure the best way to displa my original data ? github - how to get rid of the signs 1- overall national trend (use original graph) 2- i want to show the variation between different regions using healthboards 3- link it to deprevation and look at prescriptions per person 4- can i do a map that shows pre covid and post covid side by side would that count as one 5- voilin plot across differet SIMDs to compare smaller unit of data - gp practice (postcode that links to SIMD) PATCHWORK - MAPS TOGETHETE reference line to show the split between precovid, covid and postcovid
every dot is a gp practice - gp practice - dataset - adressess (assessment prep) voilin plot if messy add transperency open data use quintiles for voilin plot do a code that says if not installed install and load packages
percentage increase
overall trend map antidepressant prescribing per head by healthboard facet by winter year? boxplot by SIMD dumbell plot / lollipop graph - percentage change